WPS5170 Policy Research Working Paper 5170 Social Impacts of Climate Change in Chile A Municipal Level Analysis of the Effects of Recent and Future Climate Change on Human Development and Inequality Lykke E. Andersen Dorte Verner The World Bank Sustainable Development Department Social Development Division January 2010 Policy Research Working Paper 5170 Abstract This paper uses municipality level data to estimate the Chile, and it has contributed to reduced poverty and general relationship between climate, income, and life reduced inequality of health outcomes. expectancy in Chile. The analysis finds that incomes are Whereas temperatures in the past have shown a negatively related to temperature, while life expectancy downward trend for most of the Chilean population, is not significantly related to average temperatures. Both climate models suggest that they will increase in the incomes and life expectancy are greater in areas with future, and that there will be a reduction in precipitation either very little rain or a lot of rain. The authors use the in the central part of Chile. The analysis simulates the estimated relationships to simulate the effects of both likely effects of these projected climate changes over the past (1958­08) and future (2008­58) climate change. next 50 years. The findings suggest that expected future The findings indicate that past climate change has been climate will tend to reduce incomes across the whole favorable for the central, and most populous, part of country, with an average reduction of about 7 percent, all other things equal. This paper--a product of the Social Development Division, Sustainable Development Department--is part of a larger effort in the department to reduce poverty and vulnerability to climate change. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at dverner@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Social Impacts of Climate Change in Chile: A municipal level analysis of the effects of recent and future climate change on human development and inequality* by Lykke E. Andersena Dorte Vernerb Keywords: Climate change, social impacts, Chile. JEL classification: Q51, Q54, O15, O19, O54. * This paper forms part of the World Bank research project "Social Impacts of Climate Change and Environmental Degradation in the LAC Region." Financial support from the Danish Development Agency (DANIDA) is gratefully acknowledged. The assistance of Yyanny Cruz Aguayo is greatly appreciated as are the comments and suggestions of Kirk Hamilton, Jacoby Hanan, John Nash, Maximillian Shen, and David E. Yuravlivker. The findings, interpretations, and conclusions expressed in this paper are those of the authors and do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. a Institute for Advanced Development Studies, Calle 21 #8350, Piso 7, Calacoto, La Paz, Bolivia. Phone/Fax: +591-2-214-6069. Please direct correspondence concerning this paper to landersen@inesad.edu.bo. b The World Bank, Latin America and Caribbean Region, Washington, DC. E-mail: dverner@worldbank.org. 1. Introduction The Chilean territory stretches from the tropics to Antarctica and from the middle of the Pacific Ocean (Easter Island) to the Andean peaks almost 7 kilometers above sea level. This makes for tremendous climatic variation within a single country, and in this paper we will exploit some of this spatial variation to investigate how climate and climate change might affect human development in Chile. Climate is affecting people and their activities in many complex and intertwined ways. Some activities, such as agriculture and tourism, are more sensitive to climatic conditions than others (e.g. accounting and teaching), and some population groups, such as people with poor health and those who work outdoors, are more sensitive to climate stresses than other groups. Not only does climate change affect each individual differently, but climate change in one place can have indirect effects on people on the other side of the globe, even if they don't experience any climate change themselves.1 Despite these multi-dimensional complexities, and without understanding all the underlying mechanisms and links, it is possible to assess the likely net effects of climate change on different population groups. In order to assess how climate change is likely to affect a particular population, two things are necessary: First we have to understand how climate is currently affecting them, and second we have to understand how climate is changing. A simple way to gauge how climate affects human development is to compare human development across regions with different climates. For example, Horowitz (2006) uses a cross-section of 156 countries to estimate the relationship between temperature and income level. The overall relationship found is very strongly negative, with a 2F increase in global temperatures implying a 13 percent drop in income. This is very dramatic, but the relationship is thought to be mostly historical and thus not very relevant for the prediction of the effects of future climate change. In order to control for historical factors, the paper includes colonial mortality rates as an explanatory variable, and finds a much more limited, but still highly significant, contemporaneous effect of temperature on incomes. The contemporaneous relationship estimated implies that a 2F increase in global temperatures would cause approximately a 3.5 percent drop in world GDP. In order to further control for historical differences, Horowitz (2006) uses more homogeneous sub-samples, such as only OECD countries or only countries from the Former Soviet Union, and the negative relationship still holds. However, for directions for further research, he recommends empirical studies of income and temperature variations within large, heterogeneous countries, which would provide much more thorough control for historical differences. This is exactly what we will do in the present paper. Using data from 333 municipalities (comunas) in Chile, we will estimate short-run relationships between climate variables 1 For example, an extensive drought in Central America might damage its coffee-production and the reduced supply of coffee would increase the world price of coffee, which could benefit coffee-producers in other parts of the world. 2 (average temperature and precipitation) and income, as well as between climate variables and life expectancy. While it is always dangerous to make inferences about changes in time from cross-section estimates, these relationships can at least be used to gauge the likely direction and magnitude of effects of climate change in Chile. Two different types of climate change will be assessed. First, the documented recent climate change in each of the 333 municipalities, as estimated from average monthly temperature series from 1948 to 2008 for all the Chilean meteorological stations that have contributed systematically to the Monthly Climatic Data for the World (MCDW) publication of the US National Climatic Data Center. Second, we will use the predictions of the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC4) climate models to simulate the likely effects of projected future climate change in Chile. The rest of the paper is organized as follows. Section 2 describes the data sources and provides descriptions of the key variables. Section 3 estimates the cross-municipality relationships between climate and human development, controlling for other key variables that also affect development. Section 4 analyzes past climate change for 7 meteorological stations across Chile, and estimates average trends for each municipality. Section 5 uses the findings from sections 3 and 4 to simulate the effects of past climate change on incomes in each of the 333 municipalities in Chile. Section 6 summarizes the climate changes that are expected for Chile during the next 50 years, and section 7 simulates the likely effects of these changes on incomes. Section 8 concludes. 2. The data The data used for this paper consist of both cross-section data and time series data. The municipal level cross-section data-base, which is used to estimate the relationship between climate and development in Chile, is constructed using data from different sources. Table 1 lists the variables, their definitions, and the sources of the information. 3 Table 1: Variables in the Chilean municipal level data base Variable Definition Unit Source Total population The number of inhabitants in the - 2002 Census comuna. Urbanization rate The share of the population who - 2002 Census lives in urban areas, where urban is defined as human settlements that has more than 2000 inhabitants or that has between 1,001 and 2,000 inhabitants with more than 50% of the population economically active in secondary or tertiary activities. Years of education Average number of years of Years 2002 Census education of the population aged 15 or more. Life expectancy Life expectancy at birth for each Years UNDP Chile comuna in 1998. This variable (2000) was derived from the health index of UNDP's Human Development Index. Per-capita income Average household income per USD/month CASEN 2006 capita in each comuna, as estimated from the CASEN 2006 survey, which is representative at the comuna-level except for a few sparsely populated comunas. Pesos were converted into USD using the exchange rate 530.28 pesos/dollar. Latitude Latitude of the main city in the Degrees Google Earth, comuna. south of GeoMaker equator Longitude Longitude of the main city in the Degrees west Google Earth, comuna. of the Prime GeoMaker Meridian Elevation Elevation of the main city in the Kilometers Google Earth, comuna. above sea GeoMaker level Normal average The average annual temperature ºC Chilean annual temperature in the main city of the comuna as Meteorological measured over a reference period Department and (typically 1951-1990). www.worldclimate .com Normal annual The average annual precipitation Meters Chilean precipitation in the main city of the comuna as Meteorological measured over a reference period Department and (usually 1951-1990). www.worldclimate .com 4 As we did not have meteorological data for each and every municipality in Chile, this information was estimated. Since average annual temperature in any particular location depends principally on distance from the equator and elevation above sea-level, we estimated a simple model (see Table 2) using information on average annual temperature, latitude, and altitude for all the Chilean stations for which we could obtain "normal" temperature data (from www.worldclimate.org). In order to get sufficient variation in the elevation variable, we included 4 stations located close to Chile (Catamarca, Mendoza, Arequipa, and Oruro). Table 2: Model used to estimate temperature in Chilean municipalities with missing data Variable Coefficient t-value P-value Elevation -2.1518 -4.48 0.0000 Latitude -0.3883 -8.87 0.0000 Constant 27.2469 17.25 0.0000 No. of observations 27 R2 0.7669 The model indicates that, for every kilometer of elevation, the temperature drops 2.15 degrees Celsius, and for every latitudinal degree further south, the temperature drops 0.39 degrees. This information was used to estimate temperature in all the remaining municipalities, using the altitude and latitude of the municipality capital. Precipitation does not present such simple regularities, so in order to estimate precipitation for the municipalities where this information was missing; we extended the findings from particular stations to the macro-region to which they belong. For example, for the northernmost macro-region, Tarapacá, we have precipitation data from two stations (Arica and Iquique), so we average the precipitation for these two stations and use this average for the remaining municipalities in the macro-region. This is a reasonable process, as Chile is a long, narrow country divided into 13 relatively homogeneous macro-regions from north to south. The three northernmost macro-regions (Tarapacá, Antofagasta, and Atacama) have desert climates with hardly any rain, but rain increases towards the south to a maximum in the macro-region Aisén, and then it falls again as we move further south. As can be seen from Table 3, the variation in precipitation within each macro-region is quite small compared to the variation between macro-regions. One exception is the Aisén region, which is crossed by a mountain range, and precipitation is substantially larger on the western side of the range. Therefore all municipalities on the western side will be assigned the same precipitation as Aisén, whereas the municipalities on the eastern side will be assigned the same precipitation as Coihaique. 5 For one macro-region, O'Higgins, we did not have even one meteorological station, so we assumed that precipitation there was the average between the region immediately north of it (Metropolitan Region) and the region immediately south of it (Maule). Table 3: Rules used to infer precipitation in municipalities for which precipitation data was not available in Chile Average annual Available Average annual precipitation Macro-region meteorological precipitation for for station stations macro-region (mm) (mm) 1. Tarapacá - Arica 0.9 1.5 - Iquique 2.1 2. Antofagasta - Antofagasta 3.5 3.5 3. Atacama - Caldera 29.3 32.3 - Copiapo 11.4 - Vallenar 56.1 4. Coquimbo - La Serena 119.0 119.0 5. Valparaíso - Quintero 329.0 350.8 - Valparaíso 372.5 0. Metropolitan Region - Santiago 312.5 325.4 - Pudahual 338.2 6. O'Higgins 523.6 7. Maule - Curicó 721.7 721.7 8. Bío Bío - Concepción 1276.4 1276.4 9. Araucanía - Temuco 1152.7 1152.7 10. Los Lagos - Valdivia 2445.3 1800.0 - Osorno 1436.9 - Puerto Montt 1972.8 - Quellón 1344.8 11. Aisén - Aisén 2826.1 2826.1 (West) - Coihaique 596.1 596.1 (East) 12. Magallanes - Punta Arenas 397.4 397.4 Source: Authors' estimation based on data from www.worldclimate.org . Note: Island stations excluded due to dramatically different precipitation patterns than in mainland Chile. In order to assess the climate change trends in the different parts of Chile, we obtained monthly temperature and precipitation data from 1948 to 2008 from the Monthly Climatic Data for the World (MCDW) publication of the US National Climatic Data Center.2 Section 4 below contains a more detailed description and analysis of this data. 2 This data is available for free at http://www7.ncdc.noaa.gov/IPS/mcdw/mcdw.html (although in a very inconvenient format). 6 3. Modeling climate and human development In this section, we will estimate the contemporary relationship between climate and human development in Chile. Two dimensions of human development will be analyzed: income and health, because these are the ones that most directly could be affected by climate change. Education, another human development indicator, is treated as an explanatory variable instead of a dependent variable. As several researchers have pointed out, the relationship between temperature and development is likely to be hump-shaped, as both too cold and too hot climates may be detrimental for human development (Mendelsohn, Nordhaus and Shaw, 1994; Quiggin and Horowitz, 1999; Masters and McMillan, 2001, Tol, 2005). In order to allow for this possibility we include both average annual temperature and its square in the regression. The same argument also holds for precipitation and possibly also urbanization rates, which is why we also include precipitation and urbanization rates squared. Thus, the regressions in this section will take the following form: ln y i 1 tempi 2 tempi2 3 raini 4 raini2 5 edui 6 urbi urbi2 i where yi is a measure of the income level in municipality i, tempi and raini are normal average annual temperature and normal accumulated annual precipitation in municipality i, edui is a measure of the education level (average years of schooling), urbi is the urbanization rate of the municipality, and i is the error term for municipality i. The life expectancy regression will take the same form as the income regressions, except that we will not apply the natural logarithm to the dependent variable. All regressions are weighted OLS regressions, where the weights consist of the population size in each municipality. The regression findings for both income and life expectancy are reported in Table 4. 7 Table 4: Estimated short-term relations between climate and income/life expectancy in Chile (1) (2) Explanatory variables (log per capita income) (life expectancy) 9.4701 58.0242 Constant (248.77) (18.04) -0.0592 1.0571 Temperature (-7.37) (2.27) - -0.0560 Temperature2 (-3.12) -0.3153 -4.9969 Precipitation (-4.30) (-5.57) 0.1195 1.5325 Precipitation2 (3.68) (3.87) 0.3504 0.9710 Education level (30.11) (7.44) - -2.5520 Urbanization rate (-2.81) -0.6291 - 2 Urbanization rate (-10.52) Number of obs. 324 333 R2 0.823 0.364 Source: Authors' estimation based on assumptions explained in the text. Note: Numbers in parenthesis are t-values. When t-values are numerically larger than 2, we will consider the coefficient to be statistically significant, corresponding to a confidence level of 95%. The findings at the bottom of the table show that just these four explanatory variables (temperature, precipitation, education, and urbanization rates) explain more than 82 percent of the variation in incomes between the municipalities in Chile. This is a very good fit, which suggests that we have included the most important explanatory variables, and that including addition variables would make little difference. The same four variables only explain about 36 percent of the variation in life expectancy, which is less impressive but still good for a cross-section model. Education is by far the most important variable, explaining about 73 percent of the variation in incomes and about 26 percent of the variation in life expectancy. The remaining variables are also all statistically significant, but sometimes linearly and other times non-linearly. As it is difficult to judge the relationships directly by looking at the estimated coefficients, we have plotted the estimated relationships in Figure 1. The axes are scaled to represent the actual range of temperatures, precipitation, incomes, and life expectancies experienced in different Chilean municipalities, so that the magnitude of climate impacts can be seen in the appropriate perspective. Panel (a) shows a negative relationship between average annual temperature and per capita income, with inhabitants of the warmest regions earning less than half the income of inhabitants from the coolest regions. 8 Panel (b) shows that life expectancy in the short run (when holding other factors constant) is almost 5.1 years shorter in the warmest regions compared to the regions with the most optimal temperatures. However, the relationship between temperature and life expectancy is not very tight. Indeed, the wide confidence interval on the estimated relationship suggests that there is no statistically significant relationship between temperature and life expectancy. Figure 1: Estimated short-term relations between temperature/precipitation and income/life expectancy in Chile (a) Temperature and Income (b) Temperature and life expectancy 1715 Per capita income (USD/month) 78 1515 73 Life expectancy (years) 1315 68 1115 63 915 58 715 53 515 48 315 115 43 5 7 9 11 13 15 17 19 5 7 9 11 13 15 17 19 Average annual temperature (ºC) Average annual temperature (ºC) (c) Precipitation and Income (d) Precipitation and life expectancy 1715 Per capita income (USD/month) 78 1515 73 Life expectancy (years) 1315 68 1115 63 915 58 715 53 515 48 315 115 43 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 Average annual rainfall (m) Average annual rainfall (m) Source: Graphical representation of the estimation findings from Table 3. Note: The thick red line represents the point estimate from Table 3, whereas the thin black lines delimit the 95% confidence intervals as estimated using Stata's lincom command. Panel (c) and (d) suggest that people in Chile do better with either very little rain or with a lot of rain. For intermediate amounts of rain, both incomes and life expectancy are lower. This is somewhat counter-intuitive, but the findings are robust. 9 4. Recent climate change in Chile In this section we will analyze climate data for Chile from May 1948 to March 2008 to test whether there are any significant trends, and whether these trends differ between regions. We will use the Monthly Climatic Data for the World database collected by the National Climatic Data Center (NCDC) in the US. This project started in May 1948 with 100 selected stations spread across the World, including five in Chile. Since then, many more stations have been included in the data base, but only seven stations in Chile have contributed systematically throughout the period, with only inconsequential gaps. These are listed in Table 5, ordered from north to south. Table 5: Meteorological stations in Chile with almost complete monthly data, 1948-2008 Elevation Station Latitude Longitude (m) Arica 18°21'S 70°20'W 55 Antofagasta 23°26'S 70°27'W 140 La Serena 29°55'S 71°12'W 146 Isla Juan Fernandez 33°37'S 78°49'W 30 Valdivia 39°38'S 73°05'W 43 Puerto Montt 41°26'S 73°06'W 110 Punta Arenas 53°00'S 70°51'W 90 Source: NCDC's Monthly Climatic Data for the World. The original data was organized in 61 printed volumes with 12 issues in each (one for each month of the year), totaling 719 months. All data has been quality-checked and was published by the NCDC about three months after the raw data had been collected. From each of these monthly issues, we extracted average monthly temperature and total monthly precipitation for the seven Chilean stations listed above, in order to create time series for each station. Once the temperature and precipitation series had been constructed and checked for unrealistic values (there was only one, which was eliminated), we proceeded to calculate "normal" temperatures and "normal" precipitation for each station-month for the reference period 1960-90. Table 6 shows the average "normal" values for temperature and precipitation for each month for each of these stations. It is seen that the climate differs dramatically from region to region, with Arica and Antofagasta being located in tropical desert climates with almost no rain, while Valdivia and Puerto Montt are located in temperate, wet climates. There is also a sub-tropical island, Isla Juan Fernandez, with mild temperatures and moderate amounts of rain, and Punta Arenas with very cold winters, due to the location far south. 10 Table 6: Average temperature (ºC) and precipitation (mm) for 1960-90, by station in Chile Isla Juan Arica Antofagasta La Serena Fernandez Valdivia Puerto Montt Punta Arenas Temp. Prec. Temp. Prec. Temp. Prec. Temp. Prec. Temp. Prec. Temp. Prec. Temp. Prec. January 22.1 0.0 20.2 0.0 17.1 0.0 18.7 32.4 16.3 57.6 14.4 88.2 10.5 36.5 February 22.2 0.2 20.1 0.0 16.9 0.0 18.7 30.8 15.5 51.4 14.2 82.2 10.2 28.1 March 21.5 0.0 18.9 0.0 15.6 0.3 17.9 61.9 13.6 70.1 12.2 97.5 8.4 29.7 April 19.5 0.0 16.9 0.0 13.6 1.1 16.7 84.5 10.8 136.4 10.1 134.2 6.0 42.7 May 17.7 0.0 15.4 0.0 12.3 6.7 15.6 147.4 9.2 296.9 8.5 210.5 3.8 41.9 June 16.5 0.2 13.9 0.2 11.1 17.3 13.7 168.0 7.3 266.5 6.8 206.0 1.8 26.9 July 15.7 0.3 13.4 0.3 10.7 35.8 13.0 178.6 7.2 316.0 6.7 238.8 1.3 35.1 August 15.6 0.1 13.7 0.9 10.9 17.5 12.5 106.2 7.7 250.8 6.7 200.8 2.2 36.1 September 16.3 2.2 14.6 0.2 11.6 6.1 12.5 79.4 8.7 154.8 8.0 148.8 4.0 23.8 October 17.5 0.0 15.6 0.0 12.8 1.6 13.4 47.3 10.9 122.4 9.5 151.8 6.4 24.5 November 18.9 0.1 17.1 0.0 14.0 0.3 15.0 38.8 13.1 69.6 11.6 104.1 8.2 30.9 December 20.7 0.0 18.6 0.0 16.2 0.0 16.9 35.7 15.2 69.3 13.4 100.9 9.8 30.1 Source: Authors' estimation based on data from the NCDC's Monthly Climatic Data for the World. 4.1. Temperature trends Using the "normal" values for each station and each month, we calculate monthly anomalies for each station for the whole period (actual temperature minus normal temperature for that month). Anomalies are easier to analyze than the raw temperature and precipitation data, since the seasonal variation is eliminated through the subtraction of normal monthly temperatures. Figure 2 plots the temperature anomalies for each station. 11 Figure 2: Monthly temperature anomalies for seven meteorological stations in Chile, 1948-2008 Arica 5.0 4.0 Temperature anomaly (ºC) 3.0 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Antofagasta 4.0 3.0 Temperature anomaly (ºC) 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 12 Isla Juan Fernandez 5.0 4.0 Temperature anomaly (ºC) 3.0 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 La Serena 6.0 5.0 4.0 Temperature anomaly (ºC) 3.0 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Valdivia 5.0 4.0 3.0 Temperature anomaly (ºC) 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 -5.0 -6.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 13 Puerto Montt 6.0 5.0 4.0 Temperature anomaly (ºC) 3.0 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Punta Arenas 4.0 3.0 Temperature anomaly (ºC) 2.0 1.0 0.0 -1.0 -2.0 -3.0 -4.0 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Source: Authors' estimation based on data from the NCDC's Monthly Climatic Data for the World. Once we have the series of temperature anomalies, it is straightforward to test whether there is a significant trend. This is done simply by regressing the anomaly on a trend- variable, which has been scaled so that the coefficient can be directly interpreted as temperature change per decade in degrees Celsius. We use a confidence level of 95 percent to decide whether the trend is statistically significant, which means that the P-value should be less than 0.05 for the trend to be significant. Table 7 shows the estimated trends for each of the seven stations in Chile. Of these, two stations show significant warming since the middle of the previous century; two show no significant change; and three show significant cooling. 14 Table 7: Estimated temperature trend (ºC/decade) for seven stations in Chile Station Trend t-value P-value # of obs. Arica 0.11 4.40 0.000 648 Antofagasta 0.01 0.49 0.628 599 La Serena 0.02 0.73 0.465 624 Isla Juan Fernandez -0.12 -6.14 0.000 531 Valdivia -0.18 -5.66 0.000 547 Puerto Montt -0.24 -9.98 0.000 605 Punta Arenas -0.06 -2.54 0.011 549 Source: Authors' estimation based on data from the NCDC's Monthly Climatic Data for the World. There seems to be a clear tendency for northern stations to be warming, southern stations to be cooling, and intermediate stations to experience no change (see Figure 3). This general pattern is confirmed by the latest IPCC report (see Trenberth et al. 2007, Figure 3.9 based on Smith and Reynolds, 2005)3. Figure 3: Estimated temperature trend (ºC/decade) versus latitude in Chile 0.2 Temperature trend (ºC/decade) 0.1 0 -0.1 -0.2 -0.3 10 20 30 40 50 60 Latitude Source: Authors' estimation based on data from the NCDC's Monthly Climatic Data for the World. We will use this simple relationship to estimate the average temperature change experienced in each of the municipalities in Chile over the last 50 years. According to a simple regression, the trend is reduced by 0.067 ºC/decade for each 10 degrees we move south. This means that at the latitude of Antofagasta (23°26'S), there is no trend in temperatures, but for each 10 degrees further south, the 50-year change is reduced by 3 Falvey and Garreaud (2009) partly confirm this, but their interpretation of the data is different. They suggest cooling along the coast and warming in the mountains. However, they only use data from 1979 to 2006, and such a short period is vulnerable to spurious trends produced by natural multi-decadal cycles. 15 0.336ºC. Just south of Punta Arenas the change in temperatures over the last 50 years would be about -1ºC. At the most northern place in Chile, the temperature change over the last 50 years would be about +0.17ºC. 4.2. Precipitation trends Figure 4 plots precipitation anomalies for the same seven meteorological stations in Chile. Figure 4: Monthly precipitation anomalies (mm/month) for 7 meteorological stations in Chile, 1948-2008 Arica 30 Precipitation anomaly (mm/month) 25 20 15 10 5 0 -5 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Antofagasta 16 Precipitation anomaly (mm/month) 14 12 10 8 6 4 2 0 -2 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 16 La Serena 200 Precipitation anomaly (mm/month) 150 100 50 0 -50 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Isla Juan Fernandez 400 Precipitation anomaly (mm/month) 300 200 100 0 -100 -200 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Valdivia 600 Precipitation anomaly (mm/month) 500 400 300 200 100 0 -100 -200 -300 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 17 Puerto Montt 800 Precipitation anomaly (mm/month) 700 600 500 400 300 200 100 0 -100 -200 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Punta Arenas 200 Precipitation anomaly (mm/month) 150 100 50 0 -50 May-48 Oct-53 Abr-59 Oct-64 Mar-70 Sep-75 Mar-81 Ago-86 Feb-92 Ago-97 Feb-03 Jul-08 Source: Authors' estimation based on data from the NCDC's Monthly Climatic Data for the World. A trend analysis, like the one performed on temperatures above, reveals two mid-latitude stations with a significant negative precipitation trend, while the remaining five stations show no significant trend (see Table 8). This is consistent with the findings in the latest IPCC report (see Magrin et al. 2007, Figure 13.1). 18 Table 8: Estimated precipitation trend (mm/decade) for seven stations in Chile Station Trend t-value P-value # of obs. Arica -0.02 -0.28 0.782 632 Antofagasta 0.02 0.75 0.451 585 La Serena 0.05 0.10 0.923 618 Isla Juan Fernandez 2.26 1.75 0.080 537 Valdivia -5.01 -2.24 0.026 621 Puerto Montt -5.26 -3.18 0.002 601 Punta Arenas 0.31 0.58 0.565 595 Source: Authors' estimation based on data from the NCDC's Monthly Climatic Data for the World. The regions that have seen significant reductions in precipitation over the last 60 years, are the ones that initially (and still) received the largest amounts of precipitation, and the significant reductions are limited to the season with the heaviest rains (May to August). Valdivia, for example, used to receive about 280 mm of rain per month during the rainy season, but this amount has been going down by about 5 mm per decade over the last half century, or about 10 percent over 50 years. For the purposes of the simulations in the following section, we will assume a 10 percent reduction in precipitation over the last 50 years for all municipalities located between 35ºS and 45ºS. 5. Simulating the impact of recent climate change Since life expectancy was not found to depend significantly on temperature, and since precipitation changes are limited to a small region in the middle of the country with already abundant precipitation, we will limit the simulation analysis to the impact of climate change on average incomes. To find the impacts of recent climate change we will compare the following two scenarios: 1) Climate Change, which is the factual scenario, and 2) No Climate Change, which is the counterfactual scenario. The Climate Change temperatures are the actual temperatures in each municipality, whereas the No Climate Change temperatures are the actual temperatures minus the temperature changes experienced over the last 50 years, according to the analysis in the previous section. Similarly, Climate Change precipitation are the actual precipitation levels, while No Climate change precipitation is 10% higher in all municipalities located between 35ºS and 45ºS, as indicated by the analysis in the previous section. The ratio of Climate Change Income to No Climate Change Income can be written as: CC Yi Yi ,CC exp 1 t i ,CC 2 t i2,CC 3 ri ,CC 4 ri 2CC ^ ^ ^ ^ , Yi , NCC exp ^1 t i , NCC 2 t i2,NCC 3 ri , NCC 4 ri 2NCC ^ ^ ^ , After estimating this ratio for each municipality, it is easy to calculate the percentage change in income levels that can be attributed to climate change. 19 At the national level, the simulation indicates that incomes have increased by about 2 percent due to recent climate change. This is mostly due to the cooling of all but the most northern regions, since colder climates have been shown above to be more favorable for income generation in Chile. Figure 5 plots the estimated change in income at the municipal level against the income level each municipality would have had in the absence of recent climate change. Table 9: Effects of recent climate change on income (% change), by region in Chile Effects of Effects of Total effect of temperature changes in recent climate Region changes precipitation change 1. Tarapacá -0.9 0.0 -0.9 2. Antofagasta -0.1 0.0 -0.1 3. Atacama 0.7 0.0 0.7 4. Coquimbo 1.3 0.0 1.3 5. Valparaíso 1.8 0.0 1.8 0. Metropolitan Region 1.9 0.0 1.9 6. O'Higgins 2.1 0.0 2.1 7. Maule 2.4 0.8 3.2 8. Bío Bío 2.7 0.1 2.8 9. Araucanía 3.0 0.5 3.5 10. Los Lagos 3.5 -2.5 0.9 11. Aisén 4.4 -0.8 3.6 12. Magallanes 6.0 0.0 6.0 Total 2.1 -0.1 2.0 Source: Authors' estimations. 20 Figure 5: Estimated change in incomes due to past climate change versus initial income level, municipal level in Chile 10 8 past climate change (% change) 6 Change in income due to 4 2 0 0 500 1000 1500 2000 -2 -4 -6 -8 -10 Income per capita without climate change (USD/month) There is a slight, statistically insignificant, negative relationship, indicating that recent climate change has not had any significant effect on the income distribution, but that it has contributed slightly towards poverty reduction in all but the most northern municipalities. 6. Expected future climate change in Chile Having quantified the impacts of climate change during the last 50 years, we now turn to an assessment of the likely impacts of possible climate change during the next 50 years. For that purpose we will use the regional climate projections made by Working Group 1 for the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, which provides a comprehensive analysis based on a coordinated set of 21 Atmosphere-Ocean General Circulation Models (Christensen et al 2007). The use of several different models allows an assessment of the level of confidence with which predictions can be made. According to the model simulations reported in Christensen et al (2007), temperatures are going to increase faster in the northern part of Chile than in the southern part (see Figure 6). This thus corresponds well to the pattern observed in the past. 21 Figure 6: Temperature and precipitation changes predicted by the climate models used by IPCC 4, 1990-2090 in Latin America and the Caribbean Source: Christensen et al 2007, Figure 11.15. Since temperatures are projected to increase almost linearly over this century, it is reasonable to assume that temperature increases over the next 50 years will range from 1.5ºC in the northernmost part of Chile to 0.75ºC in the southernmost part, so we will gradually (linearly) increase the amount of warming from 0.75ºC at 54.95ºS to 1.5ºC at 17.65ºS. Most of the models used for these projections suggest that there will be a reduction in precipitation in central Chile due to a pole ward shift of the South Pacific and South Atlantic subtropical anticyclones (Christensen et al (2007, p. 896). The magnitude of reductions that can be expected in central Chile over the next 50 years appears to be around 10 percent, which is very similar to what has happened over the last 50 years, but now starting from a slightly lower base in the central region. 7. Simulating the impact of expected future change The simulated effects of future climate change on incomes are presented in Table 10. The overall reduction in incomes we might expect due to the climate change over the next 50 years is around 7 percent. The negative effects are stronger in the north, which is expected to experience more warming than the southern regions, but all regions are projected to experience a significant negative impact from climate change (holding all else constant). 22 Table 10: Effects of future climate change on income (% change), by region in Chile Effects of Effects of Total effect of temperature changes in future climate Region changes precipitation change 1. Tarapacá -8.3 0.0 -8.3 2. Antofagasta -7.9 0.0 -7.9 3. Atacama -7.4 0.0 -7.4 4. Coquimbo -7.1 0.0 -7.1 5. Valparaíso -6.8 0.0 -6.8 0. Metropolitan Region -6.8 0.0 -6.8 6. O'Higgins -6.6 0.0 -6.6 7. Maule -6.5 0.8 -5.7 8. Bío Bío -6.4 0.5 -5.9 9. Araucanía -6.2 0.7 -5.5 10. Los Lagos -5.9 -1.6 -7.5 11. Aisén -5.4 -0.6 -6.0 12. Magallanes -4.6 0.0 -4.6 Total -6.7 0.0 -6.7 Source: Authors' estimations. Figure 7 plots the estimated change in incomes for each municipality against the current level of incomes. There is a very weak, barely statistically significant, negative relationship ( = -0.12), which indicates that while future climate change is likely to have a negative effect on incomes, it is not going to contribute to increased inequality between municipalities. 23 Figure 7: Estimated change in incomes due to future climate change versus current income, municipal level in Chile 0 Change in incomes due to future projected -2 climate change (% change) -4 -6 -8 -10 -12 -14 0 500 1000 1500 2000 Current income per capita (USD/month) 8. Conclusions In this paper we first used a municipality level cross-section database to estimate the general relationship between climate and income in Chile. We found that incomes are negatively related to temperature, while life expectancy is not significantly related to average temperatures. Both income and life expectancy were higher for either very low levels of rain or high levels of rain, whereas intermediate levels of rain were less favorable. These estimated relationships were then used to simulate the effects of both past (1958- 2008) and future (2008-58) climate change. Past changes in climates were analyzed using historical data from seven meteorological stations spread across the territory, and estimating average trends for each station. It was found that trends varied systematically across the country, with northern regions having warmed over the last half century, southern regions having cooled, and central regions having experienced little or no change in temperatures. The central region did experience significant reductions in precipitation (about 10%), whereas the regions to the north and to the south experienced no changes. The consequences of these past climatic changes were then simulated using the estimated cross-section models. The findings indicate that past climate change (reductions in temperatures and reductions in precipitation) has been favorable for the central part of Chile, where the majority of the population is concentrated, whereas the sparsely populated northern regions have suffered from warming. 24 Whereas temperatures in the past have shown a downward trend for most of the Chilean population, most climate models suggest that they should increase in the future, and that they will increase faster in the northern part of Chile than in the South. The models also indicate that the reductions in precipitation observed in the central part of Chile over the last 50 years will continue during the next 50 years, amounting to a further 10 percent drop over the next 50 years. The paper simulated the likely effects of these projected climate changes, and found that expected future climate change would reduce incomes across the whole country, with an average reduction of about 7 percent. Some qualifications to these findings are in order. First, the simulations have been carried out by varying only temperature and precipitation, but holding all other factors constant. Holding everything else constant is of course not realistic. Education levels are likely to increase and the structure of the economy is likely to keep changing towards activities that are less sensitive to the climate. If the high growth rates experienced over the last 25 years (5.2% per year) continue, incomes in 2058 would be 1,261 percent higher than now if there were no climate change, and 1,177 percent higher if climate changes as projected by the IPCC models. In either case, people will be considerably richer than they are now, and their ways of living may be so different, that the climate-income relationships of today are no longer relevant. Another factor which is assumes to be constant, but which is almost certainly going to increase, is the atmospheric CO2 concentration. It seems likely that atmospheric CO2 concentration will increase from the current level of 387 ppm to somewhere between 500 and 600 ppm 50 years from now, depending on how effective Kyoto and Post-Kyoto policies are at reducing emissions. The resulting CO2 fertilization effect may increase crop productivity significantly, as indicated by almost all studies of CO2-fertilization (e.g. Allen et al 1987; Baker et al 1989; Poorter 1993; Rozema et al 1993; Wittwer 1995; Torbert et al 2004). Ignoring this beneficial effect may imply that the estimates presented in the present paper are too pessimistic. Second, people do not necessarily have to stick around as temperatures increase, as the simulations in the present paper have assumed. Internal migration could potentially reduce the costs of climate change, if people can move towards regions with more suitable climates. Chile has climates for every taste, and the fact that the population currently is concentrated in the central region with moderate climates, whereas the extremes are sparsely populated, suggests that people have been taking advantage of this possibility in the past. Currently, however, adaptation through migration is somewhat hindered by the public housing policy. Soto and Torche (2004) show that the very effective public housing subsidies in Chile have the unfortunate side effect of tying people to their geographical location and inhibiting migration. Third, this paper compares equilibrium situations before and after climate change, but ignores transition costs. Since climate changes in Chile are quite slow, especially compared to the natural variation from month to month and from place to place, such transition costs 25 are likely small, but they may include additional investments in new reservoirs and irrigation systems, as hydroelectric facilities and water supplies are affected by changes in the water flow from melting glaciers. Finally, it should be warned that the impacts found for Chile cannot be generalized to apply to other countries. The impacts of climate change differ from country to country depending on the spatial distribution of the population, the types of activities they are engaged in, and the particular patterns of climate change. References Allen, L. H. Jr., K. J. Boote, J. W. Jones, P. H. Jones, R. R. Valle, B. Acock, H. H. Rogers & R. C. Dahlman (1987) "Response of vegetation to rising carbon dioxide: Photosynthesis, biomass, and seed yield of soybean." Global Biogeochemical Cycles 1: 1-14. Baker, J.T., L. H. Allen Jr., K. J. Boote, P. Jones, & J. W. Jones (1989) "Response of soybean to air temperature and carbon dioxide concentration." Crop Science 29: 98- 105. Christensen, J.H., B. Hewitson, A. Busuioc, A. Chen, X. Gao, I. Held, R. Jones, R.K. Kolli, W.-T. Kwon, R. Laprise, V. Magaña Rueda, L. Mearns, C.G. Menéndez, J. Räisänen, A. Rinke, A. Sarr and P. Whetton (2007) "Regional Climate Projections." In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Falvey, M., and R. D. Garreaud (2009) "Regional cooling in a warming world: Recent temperature trends in the southeast Pacific and along the west coast of subtropical South America (1979­2006)" Journal of Geophysical Research, 114, doi:10.1029/2008JD010519. Horowitz, J. K. (2006) "The Income-Temperature Relationship in a Cross-Section of Countries and its Implications for Global Warming." Department of Agricultural and Resource Economics, University of Maryland, Submitted manuscript, July. http://faculty.arec.umd.edu/jhorowitz/Income-Temp-i.pdf Magrin, G., C. Gay García, D. Cruz Choque, J.C. Giménez, A.R. Moreno, G.J. Nagy, C. Nobre and A. Villamizar (2007) "Latin America. Climate Change 2007: Impacts, Adaptation and Vulnerability." Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson, Eds., Cambridge University Press, Cambridge, UK, 581-615. Masters, W. A. and M. S. McMillan (2001) "Climate and Scale in Economic Growth," Journal of Economic Growth, 6(3): 167-186. Mendelsohn, R., W. Nordhaus and D. Shaw (1994) "The Impact of Global Warming on Agriculture: A Ricardian Analysis," American Economic Review, 84(4): 753-71. Poorter, H. (1993) "Interspecific variation in the response of plants to an elevated ambient CO2 concentration." Vegetation 104/105: 77-97. 26 Rozema, J., Lambers, H., van de Geijn, S.C. and Cambridge, M.L. (eds.) (1993) CO2 and Biosphere. (Advances in Vegetation Science 14). Kluwer Academic Publishers, Dordrecht. Smith, T. M., and R. W. Reynolds (2005) "A global merged land and sea surface temperature reconstruction based on historical observations (1880­1997)" Journal of Climate 18: 2021­2036. Soto, R. and A. Torche (2004) "Spatial Inequality, Migration, and Economic Growth in Chile" Cuadernos de Economía, 41: 401-424. Tol, R. S. J. (2005) "Emission abatement versus development as strategies to reduce vulnerability to climate change: an application of FUND." Environment and Development Economics, 10: 615-629. Torbert, H. A., S. A. Prior, H. H. Rogers and G. B. Runion (2004) "Elevated atmospheric CO2 effects on N fertilization in grain sorghum and soybean" Field Crops Research, 88(1): 57-67. Trenberth, K.E., P.D. Jones, P. Ambenje, R. Bojariu, D. Easterling, A. Klein Tank, D. Parker, F. Rahimzadeh, J.A. Renwick, M. Rusticucci, B. Soden and P. Zhai (2007) "Observations: Surface and Atmospheric Climate Change." In: Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Quiggin, J. and J. K. Horowitz (1999) "The Impact of Global Warming on Agriculture: A Ricardian Analysis: Comment," American Economic Review, 89(4): 1044-45. UNDP Chile (2000) Desarrollo Humano en las Comunas de Chile. Temas de Desarrollo Humano Sostentable, Número 5. Gobierno de Chile and PNUD Chile. Wittwer, S. H. (1995) Food, Climate, and Carbon Dioxide: The Global Environment and World Food Production. CRC Press/Lewis Publishers, Boca Raton, Florida. 27